Insights by Ed Lee, MD, MPH, Chief Medical Officer, Nabla
Key Points
- Ambient AI’s deepest ROI is not minutes saved. It’s reducing cognitive burden and restoring presence in the exam room.
- Change management, not model quality, is the make-or-break factor in AI adoption.
“At the end of the day, AI is just technology. If we do this right, it shouldn’t feel technical. It should feel more human.”
Dr. Ed Lee’s thought is simple, but it’s also a standard that healthcare leaders can actually use. It reframes the discussion away from model capability and toward lived experience. What do clinicians and patients feel when AI shows up in the room?
Dr. Lee’s perspective comes from spending years in one of the most operationally disciplined care-delivery environments in the country. He grew as a practicing clinician at Kaiser Permanente, where integrated payer-provider delivery forced every workflow change to meet a high bar. Technology could not be adopted simply because it was new. It had to help clinicians focus on patients and reduce the friction that screens and administrative tasks introduced between people and care.
Today, as Chief Medical Officer at Nabla, Dr. Lee is applying those lessons to ambient AI and clinical copilots. He’s explicit that the end goal is not efficiency alone. It’s restoring joy in medicine, reducing cognitive burden, and rebuilding the patient-physician relationship. In his view, that’s where the real ROI lives.
Listen to the full conversation
The hardest part of AI adoption isn’t the AI
When asked about what he learned at Kaiser Permanente, Dr. Lee doesn’t start with feature sets or architecture. He starts with human interaction and clinicians’ ability to focus.
He describes how easily technology can unintentionally get in the way of personalized care, and how the right tools should remove friction rather than add it. But when the conversation shifts to adoption, his answer becomes even more direct: “The hard part is not the technology. It’s change management.”
Dr. Lee puts it plainly to our host, explaining that change management is often the hardest part of implementing new technologies. Clinicians are scientific and evidence-driven. They want to understand “the why.” They want proof that a new workflow will improve care, not simply create another layer of tasks.
That’s why he argues clinician involvement from day one is non-negotiable. If clinicians aren’t brought in from the beginning, he believes teams are “often destined to fail.” This isn’t a philosophical point. It’s an implementation reality. Even the best tool will stall if it’s introduced as something imposed on clinicians rather than built with them.
He also addresses a familiar tension in the current AI narrative. Yes, AI can perform well on standardized tests and sometimes generate outputs that look smarter than what humans could write quickly. But he insists the appropriate framing remains augmentation, not replacement. The clinician must integrate information into the clinical context and remain responsible for the decision.
In practice, Dr. Lee’s message is a reminder that the adoption playbook is not about persuading people that AI is amazing. It’s about building trust through evidence, involvement, and workflow fit.
Ambient AI’s ROI is cognitive relief and restored agency, not a stopwatch metric
The early story of ambient AI was mostly about time savings. And Dr. Lee acknowledges that’s how many organizations first evaluated it: minutes per encounter, hours per day, pajama time reduced.
But he points out something important. The data has evolved, and the experience varies. Some clinicians do save substantial time. Others save less per encounter than the early narrative suggested. Some still have after-hours work even with ambient tools.
And then he pivots to what he sees as the deeper value: agency.
Dr. Lee explains that ambient tools can help clinicians budget their time as they choose. A clinician could compress the day and finish earlier. Or they could invest the recovered capacity back into their patients, slowing down to develop relationships, think more carefully, and communicate more clearly. The key is that the clinician regains control over the time and attention economy of the clinical day.
From there, he connects ROI to something bigger than time: cognitive load, cognitive burden, and meaning.
He argues that the real gain is that clinicians can do what they went into medicine to do. Not to be a transcriptionist. Not to be glued to a computer. But to be a caregiver and scientist who applies evidence to improve lives.
That’s where he believes ambient AI becomes strategic. Burnout reduction and retention aren’t abstract. They are operational outcomes. The tool may not eliminate every after-hours minute, but if it restores attention and reduces mental strain, it can improve the clinician experience in a way that matters in the long term.
Dr. Lee also highlights an unexpected “win-win” effect: ambient AI can improve the patient experience. As clinicians verbalize more to ensure the technology captures the right context, patients often feel more engaged. They hear more explanations. They experience a more meaningful interaction. What began as a documentation tool can become a relationship tool, almost as a byproduct of how clinicians use it.
That’s a key point for leaders evaluating ROI. If you only measure time saved, you might miss the bigger transformation: better communication, stronger trust, and improved clinician-patient connection.
The next frontier is workflow-native intelligence
For Dr. Lee, ambient documentation is only the first layer.
He describes the next phase already arriving. It includes diagnosis capture, coding support, surfacing the right ICD-10 and CPT codes, and improved documentation to support the financial integrity of care delivery. He’s clear that accurate documentation can improve how organizations capture what they are doing and justify it appropriately.
But he also points beyond the financial layer into clinical workflow intelligence:
- chart summarization that distills decades of patient history into what matters now
- surfacing key context at the point of care
- clinical decision support that suggests diagnoses, diagnostic tests, and treatment options
- orders that can be staged on behalf of clinicians, as long as the trust layer is strong
He notes that adoption of decision support will depend on trust, which takes time to build. But he believes clinicians will be enthusiastic if the tool supports their workflow without adding noise.
This is also where his “friction” framing returns. It’s not just what the tool can do. It’s how it does it.
If a tool adds five extra clicks per patient, clinicians won’t use it. Usability and integration matter as much as capability. Dr. Lee emphasizes that “hope is not a strategy.” You can’t release tools and assume adoption happens. Adoption must be engineered through workflow understanding, usability design, and deliberate change management.
In his best-case vision, AI becomes invisible infrastructure. It fades into the workflow. It reduces friction and supports clinicians through the entire loop: intake, documentation, orders, summarization, and decision support.
Done right, AI shouldn’t feel technical. It should feel human.
Dr. Lee’s message is consistent across the episode. The goal of AI in healthcare is not to make clinicians type faster. It’s to help clinicians be more present with patients, to reduce administrative burdens, and to restore the human side of care.
He doesn’t deny the importance of ROI. He simply reframes it. The deeper ROI of ambient AI extends beyond minutes saved. It’s cognitive relief, restored agency, reduced burnout, better recruitment and retention, and more meaningful patient interactions.
And he grounds the future in execution reality. The next wave will bring decision support, diagnosis capture, and chart summarization deeper into workflows, but only if organizations do the hard work of clinician involvement and change management. Technology alone won’t carry adoption. Trust and usability will.
The Takeaway
Dr. Ed Lee’s view of healthcare AI is refreshingly grounded: the goal isn’t efficiency for its own sake, it’s restoring human connection in care. Ambient AI is the first major proof point because it reduces cognitive burden, gives clinicians agency over their time, and can improve the quality of clinician-patient communication as a natural byproduct. But Dr. Lee is clear that the hardest part isn’t the tool, it’s adoption: change management, workflow fit, and clinician involvement from day one. As ambient evolves into chart summarization, diagnosis capture, and decision support embedded directly into the clinical workflow, the standard should stay the same. Done right, AI shouldn’t feel technical. It should feel human.
Sitting at the intersection of integrated care delivery experience and real-world ambient AI deployment, Dr. Lee’s unique insights are especially valuable:
- Change management is the hardest part of AI adoption, and clinician involvement from day one is non-negotiable.
- AI outputs can be convincing, which makes clinician expertise and context essential to prevent subtle outsourcing of judgment.
- Ambient’s deepest ROI is cognitive relief and restored agency, not just time savings per encounter.
- The patient experience can improve as clinicians explain more in real time, making conversations more engaging and meaningful.
- The next frontier is workflow-native intelligence: diagnosis capture, coding support, chart summarization, and decision support at the point of care.
- Adoption depends on frictionless integration: if it adds clicks, it won’t scale, regardless of how “smart” it is.